Vision & documents

DynaMath leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-10, the highest measured score on DynaMath is 86.8% by GPT-5.2. 62 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Mythos 5	88.5%	estimated ± 0.8 pp, low confidence
2	Qwen3.8 Max	88.5%	estimated ± 0.8 pp, low confidence
3	Qwen3.8-Omni-Flash	88.2%	estimated ± 0.8 pp, low confidence
4	Kimi K3	88.2%	estimated ± 0.8 pp, low confidence
5	Claude Opus 4.7 (Adaptive)	88.2%	estimated ± 0.8 pp, low confidence
6	Qwen3.8-Flash-Next	88.1%	estimated ± 0.8 pp, low confidence
7	Qwen3.8-27B	88.0%	estimated ± 0.8 pp, low confidence
8	Claude Opus 4.8	88.0%	estimated ± 0.8 pp, low confidence
9	GLM-5.3-Flash	87.9%	estimated ± 0.8 pp, low confidence
10	Gemini 3.7 Flash	87.8%	estimated ± 0.8 pp, low confidence
11	Muse Spark 1.1	87.7%	estimated ± 0.8 pp, low confidence
12	Claude Sonnet 5	87.7%	estimated ± 0.8 pp, low confidence
13	Sakana Fugu-Ultra	87.4%	estimated ± 0.8 pp, low confidence
14	Muse Spark	87.4%	estimated ± 0.8 pp, low confidence
15	Qwen3.7 Plus	87.2%	estimated ± 0.8 pp, low confidence
16	GPT-5.6 Sol	87.2%	estimated ± 2.7 pp, low confidence
17	Seed 2.1 Pro	87.1%	estimated ± 0.8 pp, low confidence
18	Sakana Fugu	87.1%	estimated ± 0.8 pp, low confidence
19	Gemini 3.5 Flash	86.9%	estimated ± 0.8 pp, low confidence
20	GPT-5.2	86.8%	measured
21	GPT-5.5	86.6%	estimated ± 2.7 pp, low confidence
22	GPT-5.4	86.5%	estimated ± 0.8 pp, low confidence
23	Seed 2.1 Turbo	86.4%	estimated ± 0.8 pp, low confidence
24	GPT-5.6 Terra	86.4%	estimated ± 2.7 pp, medium confidence
25	Qwen3.5 397B	86.3%	measured
26	Inkling	86.3%	estimated ± 0.8 pp, medium confidence
27	Qwen3.6 Plus	86.1%	estimated ± 0.8 pp, medium confidence
28	Inkling-Small	86.1%	estimated ± 0.8 pp, medium confidence
29	MiMo-V2.5	86.0%	estimated ± 0.8 pp, medium confidence
30	Step 3.7 Flash	86.0%	estimated ± 3.3 pp, medium confidence
31	Kimi K2.6	85.8%	estimated ± 0.8 pp, medium confidence
32	Gemini 3.1 Pro	85.7%	estimated ± 0.8 pp, medium confidence
33	dots3-note Preview	85.7%	estimated ± 2.7 pp, medium confidence
34	Qwen3.6-27B	85.6%	measured
35	Qwen3.5-27B	85.5%	estimated ± 0.8 pp, medium confidence
36	GPT-5.6 Luna	85.4%	estimated ± 2.7 pp, medium confidence
37	Grok 4.3	85.3%	estimated ± 2.7 pp, medium confidence
38	MiniMax M3	85.3%	estimated ± 2.7 pp, medium confidence
39	Muse Glimmer 30B	85.3%	estimated ± 0.8 pp, medium confidence
40	Pareto 26.9	85.2%	estimated ± 2.7 pp, medium confidence
41	Gemini 3 Pro	85.1%	measured
42	Qwen3.6-35B-A3B	85.0%	estimated ± 0.8 pp, medium confidence
43	Claude Opus 4.6	84.9%	estimated ± 2.7 pp, medium confidence
44	Kimi K2.5 (Reasoning)	84.8%	estimated ± 0.8 pp, medium confidence
45	Qwen3.5-35B-A3B	84.8%	estimated ± 0.8 pp, medium confidence
46	Claude Sonnet 4.6	84.7%	estimated ± 0.8 pp, medium confidence
47	Gemma 4 31B	84.7%	estimated ± 2.7 pp, medium confidence
48	Qwen3.5-122B-A10B	84.6%	estimated ± 0.8 pp, medium confidence
49	GPT-5.4 mini	84.5%	estimated ± 2.7 pp, medium confidence
50	Kimi K2.5	84.4%	measured
51	Nemotron 3 Nano Omni 30B A3B	84.2%	estimated ± 0.8 pp, medium confidence
52	Step 5 Preview	84.2%	estimated ± 2.7 pp, medium confidence
53	Ternary Bonsai 2 27B	83.1%	estimated ± 1.2 pp, medium confidence
54	Gemma 4 26B A4B	82.8%	estimated ± 2.7 pp, medium confidence
55	Gemini 3.1 Flash-Lite	82.7%	estimated ± 0.8 pp, medium confidence
56	Interfaze Beta	80.8%	estimated ± 2.7 pp, medium confidence
57	Claude Opus 4.5	79.7%	measured
58	Gemma 4 12B	79.0%	estimated ± 2.7 pp, low confidence
59	GPT-5.4 nano	75.7%	estimated ± 2.7 pp, low confidence
60	LFM2.5-VL-3B	75.0%	estimated ± 1.2 pp, low confidence
61	Grok 4.20	72.3%	estimated ± 0.8 pp, low confidence
62	Command A+	60.7%	estimated ± 0.8 pp, low confidence
63	ZAYA1-VL-8B	60.6%	estimated ± 1.2 pp, low confidence
64	North Micro Vision Instruct	54.4%	estimated ± 1.2 pp, low confidence
65	Gemma 4 E4B	49.8%	estimated ± 2.7 pp, low confidence
66	LFM2.5-VL-450M	45.3%	estimated ± 1.2 pp, low confidence
67	Qwen2.5-VL-32B	41.4%	estimated ± 2.7 pp, low confidence
68	Gemma 4 E2B	27.0%	estimated ± 2.7 pp, low confidence

DynaMath leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General